Variables without SpecifierExpressions

[lucaswerkmeister]

In

if (exists myVar) { ... }

the ExistsCondition has a Variable with mainToken myVar, and also a specifierExpression (what would normally be = someExpression.fooBar) that boils down to the Identifier myVar as well. That is to say, the token myVar is used twice in the AST; this is currently not expected and thus leads to every subsequent token being unexpected, each with a huge stacktrace. Lovely.

(The test only tests if(exists c =test.first) { ... }, which works because there’s an actual SpecifierExpression.)

[lucaswerkmeister]

Note to future, potentially confused self: This should not be fixed in the FormattingWriter – e. g., by detecting “we’re already past this token, ignore it” – but must be fixed in the Visitor: a solution in the FormattingWriter wouldn’t work in the case of no TokenStream (see #17).

(Past token detection might still be useful for diagnostics though.)

[lucaswerkmeister]

this is currently not expected and thus leads to every subsequent token being unexpected, each with a huge stacktrace.

Actually, that’s not true. The token is not written, handleException is called, and then the visit continues normally (handleException only prints a stacktrace). Since no token was fast-forwarded over, the rest of the formatting process can continue normally.

That’s actually bad for automated testing...